It is possible that some of the exercises are not grammatically correct. Would be good to be able to correct those. One way would be to find a library. Another way to use ChatGPT or similar service.
One way would be to employ precomputed word embeddings and replace words taking into account their proximity.
The dimensionality of embeddings can be reduced for each vocabulary, because the number of words in vocabulary is much smaller (2 orders of magnitude) than the total number of words in the embedding.
It is possible that some of the exercises are not grammatically correct. Would be good to be able to correct those. One way would be to find a library. Another way to use ChatGPT or similar service.
Some libraries to check: